| ogra | waveform, so digging a bit into the KMS future, i see that v4l2_m2m seems to be pretty well advanced already but i also see that for HW decoding we will need rpivid-v4l2.dtbo loaded, should we not have dtoverlay=rpivid-v4l2 set in sync as soon as we switch from fkms to kms ? | 11:30 |
|---|---|---|
| waveform | ogra, probably on desktop (the only platform where kms is currently set), but I'm not sure how/whether it interferes with any of the legacy stuff which still works (to some extent). Another of those "more research needed" bits | 11:38 |
| ogra | yeah ... i'm starting to play with it ... prob is that you need all userspace apps patched too ... i.e. ffmpeg, kodi, vlc ... | 11:39 |
| * ogra hugs snaps ... they make that easier ... | 11:39 | |
| ogra | one other issue i have with KMS is that our pulse snap does not find HDMI devices anymore after switching to kms ... | 11:41 |
| ogra | i guess there are also some patches needed | 11:41 |
| ogra | i realize that i can't stay in the past forever though 🙂 | 11:41 |
| waveform | odd, given that pulse on the desktop image still finds the HDMI devices happily (in fact the audio is less choppy under kms than under fkms there -- it's one of the concerns I have with switching the impish desktop to fkms -- currently testing how bad it gets on different models) | 11:41 |
| waveform | and yes, the change is painful but it will result in a much more open eco-system for the pi, which can't be a bad thing long-term | 11:42 |
| ogra | funny, i never have choppy audio in ubuntu core with fkms and pulse from candidate (i do have to hack rthe tsched=0 thingie into it though) | 11:43 |
| waveform | ah, that's the thing -- with the tsched=0 hack: choppy audio under fkms (and the hack does fix it), but kms is not choppy (on the stock impish desktop) *without* the tsched=0 hack | 11:43 |
| ogra | definitely ... but i need to give template code to our customers so i need to understand how it works (at least on the surface) | 11:44 |
| waveform | urgh, I meant *without* the tsched=0 hack -ooops | 11:44 |
| ogra | well, i see more audio devices in /proc/asound/cards when switching (there seem to be two upstream kernel devices too) but pulse (in the snap) only sees the headphone jack | 11:46 |
| waveform | hmmm, sounds like a confinement issue | 11:46 |
| ogra | not sure ... i'll try switching it to devmode for a test ... | 11:47 |
| ogra | (i wouldnt know why the confinement wouldnt make it find devices it usually has access to ) | 11:47 |
| waveform | presumably because under kms they're "different devices" to those under fkms (different kernel driver, resulting in a different device? Though ALSA/pulse may label them the same, the underlying path is presumably different) | 11:48 |
| ogra | right ... i suspect it is more on the alsa level than pulse actually | 11:49 |
| ogra | with kms enabled i have vc4hdmi and vc4hdmi1 ... and headphones ... pulse doesnt see the vc4hdmi ones | 11:49 |
| ogra | nope, devmode doesnt seem to change anything ... so not confinement ... | 11:54 |
| ogra | well, at least the headphone jack works ... (not helpful for TV audio indeed, but pulse itself seems to work) | 11:55 |
| ogra | i see pulse shipd the whole of /usr/share/alsa in the snap, do we have any special handling in the alsa configs for teh pi that 20.04 didnt have ? | 11:58 |
| waveform | not that I'm aware of, but alsa is typically desktop team so perhaps they've added some I don't know about? | 12:00 |
| ogra | i cant imagine that ... | 12:01 |
| ogra | they dont go out on their own to fiddle with pi stuff i think 🙂 | 12:01 |
| waveform | not generally, no -- still if it's not confinement it's worth a look | 12:01 |
| ogra | yeah ... all the bits on UC are 20.04 though ... so any fixes that came in later will not be included yet | 12:02 |
| waveform | good point | 12:03 |
| ogra | https://git.alsa-project.org/?p=alsa-lib.git;a=blob;f=src/conf/cards/vc4-hdmi.conf;h=027804a145d1ed9e4fe33dbb5fa7ef278230c328;hb=HEAD | 12:05 |
| ogra | aha ! | 12:05 |
| ogra | so there is a special alsa config | 12:05 |
| ogra | YAY ! | 12:22 |
| ogra | so re-packing pulse with that file and manually setting the default sink to HDMI works ... | 12:22 |
| ogra | (now, how do i make it default to HDMI instead of headphones | 12:23 |
| ogra | ) | 12:23 |
| ogra | waveform, and you are correct, no tsched hack needed with that | 12:25 |
| ogra | next stop ... roll an ffmpeg snap with v4l2-m2m support ... | 12:26 |
| waveform | ah, a special config but upstream -- that makes sense | 12:26 |
| * ogra looks at https://github.com/popcornmix/FFmpeg/tree/deinterlace | 12:26 | |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!